Creating Your First Agent

Step-by-step guide to creating and configuring a high-quality AI agent

Creating Your First Agent#

This guide walks you through creating and configuring a high-quality AI agent in AlonChat.

What is an Agent?#

An agent is your AI chatbot. Each agent has its own:

• Knowledge base (files, text, Q&A, websites, Google Drive, conversations, structured data)
• AI model (multiple providers and tiers available)
• System prompt (personality and instructions)
• Messaging channels (Chat Widget, Facebook Messenger, Instagram, WhatsApp, Telegram, Email, Help Page)
• AI Actions (Google Calendar, Google Sheets, Google Drive, Email, Leads, Payments, Web Search)
• Webhooks (send events to external systems)
• CRM (auto-classified contacts, lead scoring, follow-ups, broadcasts)

Step-by-Step: Creating an Agent#

1. Navigate to Your Project#

From the dashboard:

1. Select your project (or create one if you haven't)
2. Click Agents in the sidebar
3. Click Create New Agent

2. Configure Basic Settings#

Agent Name

• Enter a clear, descriptive name
• Examples: "Customer Support", "Sales Assistant", "FAQ Bot"
• This is for your reference -- users don't see it

Initial Message

• The first message users see when they start chatting
• Keep it friendly and clear
• Examples:
  • "Hi! I'm here to help with your questions about [Product]. What can I help you with today?"
  • "Hello! Ask me anything about our services!"
  • "Kumusta! How can I assist you? (English/Tagalog ko ay ok!)"

Initial Questions (Optional)

• Add 2-4 suggested questions to help users get started
• Examples:
  • "What are your pricing plans?"
  • "How do I reset my password?"
  • "What are your business hours?"

3. Choose an AI Model#

AlonChat supports multiple AI providers. Models and credit costs are managed in your dashboard and updated regularly:

Check the model selector in your dashboard for current available models and pricing.

Recommendation:

• Start with a budget model (1 credit) -- Great quality for most use cases
• Upgrade to premium when you need more nuanced or complex responses

4. Configure Advanced Settings (Optional)#

Click Advanced Settings to customize:

System Prompt

• Instructions that define your agent's personality and behavior
• Default works well, but you can customize for specific use cases
• Example custom prompt:
  CodeCopy

  You are a friendly customer support agent for [Company Name].
  Always be polite, concise, and helpful. If you don't know
  the answer, say "I don't have that information right now,
  but I'll connect you with a human agent who can help."
  

Temperature (0.0 - 1.0)

• Controls creativity vs. consistency
• 0.0-0.3: Very consistent, factual (recommended for customer support)
• 0.4-0.7: Balanced (recommended for general use)
• 0.8-1.0: Creative, varied responses (not recommended for factual accuracy)

Max Tokens

• Maximum length of responses
• Default: 500 tokens (~375 words)
• Increase for longer, detailed responses
• Decrease for concise answers

Context Window

• How many previous messages to include in context
• Default: 10 messages
• Higher = better context retention, but higher cost

5. Create the Agent#

Click Create Agent to finish setup.

Next Steps#

After creating your agent:

1. Add knowledge sources -- Your agent needs content to learn from
2. Train your agent -- Process the knowledge base
3. Test in the chat playground -- Verify it works
4. Set up integrations -- Connect to Facebook, Instagram, WhatsApp, Telegram, and more

Best Practices#

Do's#

• Start simple: Create a basic agent, test it, then add complexity
• Use clear initial messages: Tell users what your agent can help with
• Choose the right model: Budget models (1 credit) handle most use cases well; upgrade for complex needs
• Keep system prompts focused: Clear, specific instructions work best
• Test thoroughly: Chat with your agent before deploying to live channels

Don'ts#

• Don't skip testing: Always test before connecting to live channels
• Don't use high temperature for factual Q&A: Stick to 0.0-0.3 for accuracy
• Don't create multiple agents for the same purpose: Use one agent with comprehensive knowledge
• Don't forget to train: Adding sources doesn't automatically train the agent

Common Issues#

"My agent gives generic responses"

• You haven't added knowledge sources yet -- Add sources
• You haven't trained the agent -- Click Train Agent
• Your sources lack specific information -- Add more detailed content

"Training is taking too long"

• Large files take longer to process
• Website crawling can take time depending on page count
• Check the Training Status indicator for progress

"My agent isn't using my knowledge"

• Make sure you clicked Train Agent after adding sources
• Check that sources are in "Ready" status (not "Processing" or "Failed")
• Verify your knowledge is relevant to the questions being asked

Example: Customer Support Agent#

Here's a complete example setup:

Agent Name: Customer Support Assistant
Initial Message: Hi! I'm here to help with questions about our
  products, pricing, and support. What can I help you with today?

Initial Questions:
- What are your pricing plans?
- How do I contact support?
- What's your refund policy?

AI Model: Mid-tier model (balanced quality and cost)
Temperature: 0.2 (factual, consistent)
Max Tokens: 500

System Prompt:
You are a helpful customer support agent for Acme Corp.
Answer questions using the knowledge base provided.
Be friendly, concise, and accurate. If you don't know
the answer, say "I don't have that information. Let me
connect you with a human agent who can help!"

Sources:
- Text: Pricing information
- Q&A: Common customer questions (50 pairs)
- Website: https://acmecorp.com/docs
- PDF: Product manual

This setup creates a professional, accurate support agent that handles most common questions automatically.